Testing of a Predictive Control Strategy for Balancing Renewable Sources in a Microgrid
Mattia Marinelli; [email protected]
Center for Electric Power and Energy
DTU Risø Campus
University of California – Santa Cruz
6th Aug 2014
Summer School US-DK
2 DTU Electrical Engineering, Technical University of Denmark
Outline – I part (storage control strategy and experimental validation)
I. Introduction
II. Microgrid and energy management details
Problem definition
System layout and forecast input
Energy management strategy and microgrid components
Storage sizing and controller sensitivity
III. Experimental testing results and procedures
Experimental results on 3rd July
Experimental results on 7th July
IV. Conclusions and future developments
3 DTU Electrical Engineering, Technical University of Denmark
Introduction
• In several member countries of the European Union, small scale feed-in systems have
increased in popularity due to favorable regulations.
• Due to the fact that the conventional power plants are displaced and thus eventually shut
down, DER plants (including PV and wind) will be required to provide a predictable
production plan and to be able to grant it, even if the meteorological conditions differ
from the forecasted ones.
• It means that every producer will have to provide a day-ahead production plan
with hourly resolution, which is supposed to be granted within a given confidence
interval.
• In such setup energy storage (or demand response) can help in meeting the hourly
production plan.
4 DTU Electrical Engineering, Technical University of Denmark
Microgrid and energy management details Problem definition - hypothesis
• The day-ahead, an hourly energy production plan for the microgrid system
is defined. This plan is calculated by knowing the PV module layout information,
the WT power curve and the weather forecast.
• During the operation day the hourly production plan must be respected within
±1%.
• The storage system can be used to correct the deviations from the plan, but
it is very crucial not to overuse it, because any charge/discharge cycle would
lead to energy losses.
5 DTU Electrical Engineering, Technical University of Denmark
Microgrid and energy management details System layout in SYSLAB facility
10 kW PV Plant:
TF & Poly
11 kW WT Plant:
passive stall turbine
15 kW – 190 kWh:
Vanadium Redox Flow
PCC
Power Transit
Energy Manager
Storage
Power Ref
Estimated Production
DTU Meteo Forecast:
Solar Irr; Temp; Wind Hourly Data Microgrid Component
Simulated Models
400 V 3p
Main Network
Control System
Energy Plan
• SYSLAB facility and used subset of power
components. SYSLAB is a small-scale power
system consisting of real power components
interconnected by a three phase 400V power
grid, and some communication and control
nodes interconnected by a dedicated network,
all distributed over the Risø Campus.
6 DTU Electrical Engineering, Technical University of Denmark
Microgrid and energy management details Forecast input
• Forecasts are given by the Wind Energy Dep. two times per day for the following 48 hours:
– @ 9 am the 48 hourly series starting at 12 GMT
– @ 9 pm the one starting at 24 GMT.
-7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 70
0.05
0.1
0.15
0.2
0.25Wind speed forecast error histogram
Rela
tiv
e f
req
uen
cy
(p
u)
Wind speed error (m/s)
Mean
±STD
-0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.70
0.05
0.1
0.15
0.2
0.25
Solar irradiance forecast error histogram
Rela
tiv
e f
req
uen
cy
(p
u)
Solar irradiance error (kW/m^2)
Mean
±STD
• 6 months of comparison between the
forecasted values of wind speed and
solar irradiance and the respective
historical data are reported.
• The mean values are respectively
equal to 1.0 m/s and 34 W/m2
• The standard deviations are 2.9 m/s
and 258 W/m2
7 DTU Electrical Engineering, Technical University of Denmark
Microgrid and energy management details Microgrid components models
• In order to evaluate the PV and the WT production, proper models have been developed in Simulink.
• The hourly weather data is used as input for the models
• The PV and WT estimated hourly productions are used for defining the energy production
plan (the one that the microgrid owner is supposed to follow with a precision of ±1% during the day
of operation).
8 DTU Electrical Engineering, Technical University of Denmark
Microgrid and energy management details Energy Management Strategy
• The hourly energy plan is used by the energy manager to build the Energy Ref curve.
• The red and the green dashed curves are the control band. Whenever the energy profile
exceeds the upper or the lower bound (red and green line) the battery is activated:
– the more the distance from the objective value, the deeper the charge/discharge required.
– the opening of the lines, that means the distance between the upper and the lower band with
reference to the “Energy Ref” curve, determines the stiffness of the control
0 5 10 15 20 25 30 35 40 45 50 55 60-2
024
68
1012
Energy Plan - Microgrid
En
erg
y (
kW
h)
Time (min)
Upper Band
Lower Band
Energy Ref
9 DTU Electrical Engineering, Technical University of Denmark
Microgrid and energy management details a possible way to address the storage control – power smoothing strategy
• A storage (15 kW-190 kWh) plus wind turbine (11 kW) system aiming at smoothing the power
output at the PCC with 1 second communication delay!
0 5 10 15 20 25 30 35 40 45 50 55 60-12
-8
-4
0
4
8
12Battery (black) and Wind Turbine (blue) Powers
Po
we
r (k
W)
Time (min)
0 5 10 15 20 25 30 35 40 45 50 55 6002468
10121416
PCC (blue) and Reference (red) Powers
Po
we
r (k
W)
Time (min)
10 DTU Electrical Engineering, Technical University of Denmark
Microgrid and energy management details a possible way to address the storage control – power smoothing strategy
• …and now with 8 seconds communication delay!!! Very far from being smoothed!
0 5 10 15 20 25 30 35 40 45 50 55 60-12
-8
-4
0
4
8
12Battery (black) and Wind Turbine (blue) Powers
Po
we
r (k
W)
Time (min)
0 5 10 15 20 25 30 35 40 45 50 55 6002468
10121416
PCC (blue) and Reference (red) Powers
Po
we
r (k
W)
Time (min)
11 DTU Electrical Engineering, Technical University of Denmark
Microgrid and energy management details Concerning the power and the energy sizing of the storage
• The design of the control bands has an impact on both storage power and energy size. The two main
variables that affect the sizing of the storage are the forecast precision and accuracy (i.e. if they have
an offset and how much they span).
• Power sizing:
– In the considered setup the control bands are set to 10%, meaning that the lower band (the green line)
crosses at the 6th minute the time axis.
– This means that a storage inverter of 5 kW should be able, in case of null production of the microgrid, to
release full power for the remaining 54 minutes, leading to a maximum energy of 4.5 kWh.
– Considering a microgrid maximum hourly production of 21 kWh, which means full power output from
both PV and WT, with this control band setup, the storage is able to compensate production errors up to
21% of the microgrid hourly production.
• Energy sizing:
– The sizing of the storage capacity depends on the coherence of the forecast errors. It can be noted that
in case of alternate hourly errors, the storage alternatively charges and discharges so that the depletion
of the state of charge depends mainly on the storage internal inefficiencies.
– Therefore the storage energy rating could be relative small. If the error has the same sign for the whole
day, then the storage energy sizing criteria needs to be revisited.
12 DTU Electrical Engineering, Technical University of Denmark
Microgrid and energy management details Controller sensitivity • The relative energy error is the difference between the system energy state and the control bands. Two energy errors can
be identified depending on the crossing of the upper or the lower band.
– eup is the difference between the system energy state and the upper control band, divided by the upper control
band value. This error triggers a storage charge.
– edown is the difference between the system energy state and the lower control band, divided by the lower control
band value. This error triggers a storage discharge.
• The sensitivity of the controller is chosen in order to have the storage to store/release the maximum power, Pmax, if the
relative energy error, e, is equal or greater to 1%.
• If the error amplitude is within 0% and 1%, the storage power set-point is changed by five discrete steps 20% amplitude.
Upper relative energy
error %
Storage reference
power
Lower relative energy
error %
Storage reference
power
eup ≤ 0% 0 edown ≤ 0% 0
0 < eup ≤ 0.2 0 0 < edown ≤ 0.2 0
0.2 < eup ≤ 0.4 -20% Pmax 0.2 < edown ≤ 0.4 20% Pmax
0.4 < eup ≤ 0.6 -40% Pmax 0.4 < edown ≤ 0.6 40% Pmax
0.6 < eup ≤ 0.8 -60% Pmax 0.6 < edown ≤ 0.8 60% Pmax
0.8 < eup ≤ 1 -80% Pmax 0.8 < edown ≤ 1 80% Pmax
eup ≥ 1% -Pmax edown ≥ 1% Pmax
13 DTU Electrical Engineering, Technical University of Denmark
Experimental testing Procedure
• The testing days reported are the 3rd July 2013, which was a windy day with frequent clouds
passages and the 7th July, which was a sunny day with few wind.
• The experimental process is here explained:
1. The meteorological forecast data (solar irradiance, wind speed, air temperature) of the
studied day are used to evaluate the PV and WT output. The day-ahead forecasts that mean
the forecasts given at the 9 am of the 2nd July 2013 (for the experiment run on the 3rd July)
are taken in consideration.
2. The PV and WT model outputs are used to build the forecasted energy plan for the studied
day (3rd July).
3. The microgrid setup is prepared and the pc with the Simulink controller, aimed at managing
the storage system, is connected to the SYSLAB facility and the experiment runs.
14 DTU Electrical Engineering, Technical University of Denmark
Experimental testing test on July 3rd
0 30 60 90 120 150 180
02468
101214
Energy State and Plan
En
erg
y (
kW
h)
Time (min)
0 30 60 90 120 150 180-6
-4
-2
0
2
4
6Storage Power
Po
wer
(kW
)
Time (min)
Energy State
Upper Band
Lower Band
Energy Ref
0 20 40 60 80 100 120 140 160 1800
1
2
3
4
5Relative Energy Error (5% limited)
En
erg
y E
rro
r (%
)
Time (min)
• The experiment starts at 13:00 GMT (15:00
local time) and lasts for 3 hours.
• The overall energy plan, that means the expected
production for PV and WT, for the 3 hours
experiment is shown in the first plot. The
microgrid energy production can be also observed
in this plot (black curve).
• The storage, whose output is depicted in the
second plot, is able to compensate the deviations
from the predicted production.
• The relative deviation from the expected
production, which means the difference between
the production and the forecasts, referred to the
energy plan is reported in the third plot.
15 DTU Electrical Engineering, Technical University of Denmark
Experimental testing test on July 3rd – detail
120 125 130 135 140 145 150 155 160 165 170 175 180-202468
1012
Energy State and PlanE
nerg
y (
kW
h)
Time (min)
Energy State
Upper Band
Lower Band
Energy Ref
120 125 130 135 140 145 150 155 160 165 170 175 180-6
-4
-2
0
2
4
6Storage Power
Po
wer
(kW
)
Time (min)
16 DTU Electrical Engineering, Technical University of Denmark
Experimental testing test on July 3rd
• The power profiles of the different
components can be observed in the first
plot.
• While the overall microgrid production,
that means the power transit at the PCC, is
reported in the second plot.
0 20 40 60 80 100 120 140 160 180-5
-4
-3
-2
-1
0
1
2
3
4
5
6
7
8
9
10
11
12PV, WT and Storage outputs
Po
wer
(kW
)
Time (min)
PV
WT
Storage
0 20 40 60 80 100 120 140 160 18002468
1012141618
Microgrid PCC power transit
Po
wer
(kW
)
Time (min)
17 DTU Electrical Engineering, Technical University of Denmark
Experimental testing test results on July 3rd
3rd July – 13:00-16:00 GMT
GMT hours 13-14 14-15 15-16 13-16
PV+WT forecasted (kWh) 7.85 12.61 9.61 30.07
PV+WT produced (kWh) 8.49 10.75 11.35 30.58
PV+WT error (kWh) 0.64 -1.86 1.74 *4.24
Storage usage (kWh) -0.51 2.48 -4.10 *7.09
PCC energy transit (kWh) 7.88 12.60 9.61 30.08
PCC energy absolute error (kWh) 0.027 -0.014 -0.003 *0.044
PCC energy relative error (%) 0.35 -0.11 -0.03 **0.15
• PV+WT forecasted is the forecasted hourly production, it is the value of the energy plan at the end of the hour.
• PV+WT produced is the hourly energy produced by the renewable sources: photovoltaic and wind turbine.
• PV+WT error is the difference between energy produced and forecasted. The error is positive if there is an excess of
production.
• Storage usage is the total work of the storage that means the sum of charge and discharge regardless of the sign.
• PCC energy transit is the hourly production of the whole microgrid, thus PV, WT and storage and including the cable losses
(almost negligible due to their size).
• PCC energy absolute error is the difference between the production of the microgrid and the forecasted one.
• PCC energy relative error is the PCC energy error relative to PV+WT Forecasted.
• *calculated as the sum of the hourly errors regardless of sign;
• ** calculated as the weighted average of the hourly errors regardless of sign
18 DTU Electrical Engineering, Technical University of Denmark
Experimental testing test on July 7th
• The experiment starts at 9:00 GMT (11:00
local time) and lasts for 6 h.
• The overall energy plan, that means the
expected production for PV and WT, for the 3
hours experiment is shown in the first plot.
The microgrid energy production can be also
observed in this plot (black curve).
• The storage, whose output is depicted in the
second plot, is able to compensate the
deviations from the predicted production.
• The relative deviation from the expected
production, which means the difference
between the production and the forecasts,
referred to the energy plan is reported in the
third plot.
0 30 60 90 120 150 180 210 240 270 300 330 360
02468
101214
Energy State and Plan
En
erg
y (
kW
h)
Time (min)
Energy State
Upper Band
Lower Band
Energy Ref
0 30 60 90 120 150 180 210 240 270 300 330 360-6
-4
-2
0
2
4
6Storage Power
Po
wer
(kW
)
Time (min)
19 DTU Electrical Engineering, Technical University of Denmark
Experimental testing test on July 7th
0 30 60 90 120 150 180 210 240 270 300 330 360-5
-4
-3
-2
-1
01
2
3
4
5
67
8
9
10
1112
PV, WT and Storage outputs
Po
wer
(kW
)
Time (min)
PV
WT
Storage 0 30 60 90 120 150 180 210 240 270 300 330 36002468
1012141618
Microgrid PCC power transit
Po
wer
(kW
)
Time (min)
0 30 60 90 120 150 180 210 240 270 300 330 36002468
1012141618
PV and WT power production
Po
wer
(kW
)
Time (min)
• It is straightforward to note that the forecasts were
rather optimistic; however the storage succeeds in
managing the microgrid energy production.
• the microgrid power output smoothing is not
pursued.
20 DTU Electrical Engineering, Technical University of Denmark
Experimental testing test results on July 7th • PV+WT forecasted is the forecasted hourly production, it is the value of the energy plan at the end of the hour.
• PV+WT produced is the hourly energy produced by the renewable sources: photovoltaic and wind turbine.
• PV+WT error is the difference between energy produced and forecasted. The error is positive if there is an excess of
production.
• Storage usage is the total work of the storage that means the sum of charge and discharge regardless of the sign.
• PCC energy transit is the hourly production of the whole microgrid, thus PV, WT and storage and including the cable losses
(almost negligible due to their size).
• PCC energy absolute error is the difference between the production of the microgrid and the forecasted one.
• PCC energy relative error is the PCC energy error relative to PV+WT Forecasted.
• *calculated as the sum of the hourly errors regardless of sign;
• ** calculated as the weighted average of the hourly errors regardless of sign
7th July – 9:00-15:00 GMT
GMT hours 9-10 10-11 11-12 12-13 13-14 14-15 9-15
PV+WT forecasted (kWh) 11.56 12.78 12.82 12.40 11.46 9.67 70.69
PV+WT produced (kWh) 8.90 9.78 9.41 9.42 9.58 6.93 54.02
PV+WT error (kWh) -2.66 -3.00 -3.41 -2.98 -1.88 -2.73 *-16.66
Storage usage (kWh) 2.56 5.36 8.63 11.56 13.36 16.00 57.47
PCC energy transit (kWh) 11.48 12.60 12.70 12.38 11.38 9.57 70.11
PCC energy absolute error (kWh) -0.08 -0.18 -0.12 -0.03 -0.07 -0.10 *-0.58
PCC energy relative error (%) -0.71 -1.41 -0.90 -0.22 -0.60 -0.99 **0.80
21 DTU Electrical Engineering, Technical University of Denmark
Conclusions… • The management of the energy production of energy systems with renewable generation sets will
get more and more important in the near future and the correct prediction of the expected
production will be crucial in order to be able to manage the overall system resources.
• Improving the forecast techniques is of great importance; however errors cannot be totally
avoided. The compensation of these errors will be critical and a proper management of the
controllable resources, such as storage systems or controllable loads, is essential.
• This presentation showed a management strategy capable of controlling a storage system (limited
to 5 kW) in order to grant the forecasted energy production plan of a microgrid, formed by a 10
kW PV and an 11 kW WT.
• The overall idea consists in, by knowing the meteorological forecast for the next 24 h, controlling
the storage system in order to compensate the hourly deviations from the day-ahead energy
plan, minimizing the storage usage.
22 DTU Electrical Engineering, Technical University of Denmark
…and future developments • A first evaluation of the correlation between the meteorological forecasts errors and the sizing in
term of power and energy of the storage is analyzed even if deeper analyses are needed.
• The range of experiments will be extended across the days, taking also advantage of the fact that
the whole system can be studied in the simulated environment.
• Finally, further analysis will aim at extending the described strategy in a larger microgrid setup,
including building appliances, such as heating systems and water heaters, and electric vehicles.
23 DTU Electrical Engineering, Technical University of Denmark
Outline – II part (simulations over 1 day and storage sizing)
I. Introduction
II. Model Description
– Problem definition
– System layout and management strategy analysis
III. Simulation Results
– Simulation procedure
– Scenarios analysed (different storage power size)
IV. Conclusions and Future Developments
24 DTU Electrical Engineering, Technical University of Denmark
Model Description System layout – components are simulated
10 kW PV Plant: TF & Poly
VRB storage system
PCC
Power Transit
Energy Manager
Storage
Power Ref
Estimated Production
Meteo Forecast:
Solar Irr; Temp; Wind Hourly Data PV Plant Model
400 V 3p
Main Network
Control System
Energy Plan
25 DTU Electrical Engineering, Technical University of Denmark
Model Description Photovoltaic model
• The DC power produced by the module mainly
depends on the incident solar radiation and on
the temperature, which for instance is function
of air temperature, wind speed and radiation
itself.
• The dependence of the panel output with
different sunlight intensity and the dependence
in function of the temperature have been
evaluated in order to evaluate the reduction
from the nominal efficiency, having taken in
account that the nominal data are provided for
standard meteorological conditions (1000 W/m2
and 25 °C)
4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 200
12345
6789
10PV Production 10 kW - 1 minute sampled
Ou
tpu
t (k
W)
Time (h)
Model Output
Historical Data
4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 200
12345
6789
10PV Production 10 kW - 60 minute sampled
Ou
tpu
t (k
W)
Time (h)
Model Output
Historical Data
26 DTU Electrical Engineering, Technical University of Denmark
Model Description storage model
• The modelled dynamics regard the State-of-Charge (SOC) behaviour, the electrochemical conversion
and the thermal characterization.
• The main state variables are therefore the state of charge, the temperature and the
voltage: all the characteristic elements of the storage system (as open circuit voltage, internal
resistances, limitations and protections thresholds) present some kind of dependence from these
state variables.
27 DTU Electrical Engineering, Technical University of Denmark
Model Description Energy Management Strategy
• The input of the controller is the power measured
at the PCC, which is not directly used in the control
loop; but it is used to compute the correspondent
energy reference within the hour.
0 5 10 15 20 25 30 35 40 45 50 55 60-1
0
1
2
3
4
5
6
7
8Energy Plan - Microgrid
En
erg
y (
kW
h)
Time (min)
Upper Band
Lower Band
Energy Ref
6 6.25 6.5 6.75 7-0.5
0
0.5
1
1.5
2
2.5
3Energy Plan
En
erg
y (
kW
h)
Time (h)
Energy State
Upper Band
Lower Band
Energy Ref
6 6.25 6.5 6.75 7-1.5
-1
-0.5
0
0.5
1
1.5Battery Power
Po
wer
(k
W)
Time (h)
Ref Power
Power Output
• The red and the green lines form the control
band: whenever the energy profile exceeds the
upper or the lower bound (red and green line)
the battery is activated:
– the more the distance from the objective
value, the deeper the charge/discharge
required.
28 DTU Electrical Engineering, Technical University of Denmark
Simulation Results Simulation Procedure
• The simulation day chosen is the 4th May 2013, which was a sunny day with
frequent clouds passages.
• The simulation process follows:
1. The forecast meteorological data (solar irradiation, wind speed, air
temperature) are used to evaluate the PV output. The day-ahead forecasts
(that mean the forecasts given at the 8 am of the 3rd May 2013) are taken in
consideration.
2. The PV model output is used to build the energy plan for the studied day
3. Since the sensitivity analysis aim at finding the suitable size of the storage
system, several simulations are run using the historical production data of
the PV installed in the SYSLAB and using the Simulink validated model of the
storage.
29 DTU Electrical Engineering, Technical University of Denmark
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19-1
-0.5
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
5.5
6
6.5
7
7.5
8
8.5
9Energy Plan
En
erg
y (
kW
h)
Time (h)
Upper Band
Lower Band
Energy Ref
Simulation Results Simulation Procedure
4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 200
1
2
3
4
5
6
7
8
9
10PV Production 10 kW - 1 minute sampled
Ou
tpu
t (k
W)
Time (h)
Forecast Output
Historical Data
4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 200
1
2
3
4
5
6
7
8
9
10PV Production 10 kW - 60 minute sampled
Ou
tpu
t (k
W)
Time (h)
Forecast Output
Historical Data
PV Forecasted profile (red)
PV historical data (blue)
The total energy produced by
the PV for the studied day was
64.6 kWh, while, according to
the forecasts, it would have
been 71.8 kWh (+11%).
30 DTU Electrical Engineering, Technical University of Denmark
Simulation Results Base case – no storage
6 7 8 9 10 11 12 13 14 15 16 17 180
1
2
3
4
5
6
7
8
9System Hourly Energy
En
erg
y (
kW
h)
Time (h)
Hourly System Energy
Hourly Energy Error
6 7 8 9 10 11 12 13 14 15 16 17 180
1
2
3
4
5
6
7
8Hourly Energy Relative Error
Err
or
(%)
Time (h)
6 7 8 9 10 11 12 13 14 15 16 17 18-1
0
1
2
3
4
5
6
7
8
9Energy Plan
En
erg
y (
kW
h)
Time (h)
Energy State
Upper Band
Lower Band
Energy Ref
The relative average error
is equal to 17.3% and the
maximum one is greater
than 50%.
31 DTU Electrical Engineering, Technical University of Denmark
Simulation Results Different storage power size (with reference to the PV installed capacity)
Results with the storage size
of 3 kW (30% scenario)
6 7 8 9 10 11 12 13 14 15 16 17 18-1
0
1
2
3
4
5
6
7
8
9Energy Plan
En
erg
y (
kW
h)
Time (h)
Energy State
Upper Band
Lower Band
Energy Ref
6 7 8 9 10 11 12 13 14 15 16 17 18-3
-2.5-2
-1.5-1
-0.50
0.51
1.52
2.53
Battery Power
Po
wer
(k
W)
Time (h)
Ref Power
Power Output
6 7 8 9 10 11 12 13 14 15 16 17 180
1
2
3
4
5
6
7
8
9System Hourly Energy
En
erg
y (
kW
h)
Time (h)
Hourly System Energy
Hourly Energy Error
6 7 8 9 10 11 12 13 14 15 16 17 180
1
2
3
4
5
6
7
8Hourly Energy Relative Error
Err
or
(%)
Time (h)
32 DTU Electrical Engineering, Technical University of Denmark
Simulation Results Different storage power size (with reference to the PV installed capacity)
• It is interesting to observe the
behavior of the power measure at the
PCC (first plot) which for certain
aspects is worsen because of the
power steps which can be induced by
the storage operation.
• However it has to be kept in mind that
the objective of the control strategy
proposed is driven from an energy
perspective due to the need to respect
the scheduled energy plan.
5 6 7 8 9 10 11 12 13 14 15 16 17 18-10123456789
1011
PCC (Battery plus PV) transit
Po
wer
(k
W)
Time (h)
5 6 7 8 9 10 11 12 13 14 15 16 17 18-10123456789
1011
PV PowerP
ow
er (
kW
)
Time (h)
33 DTU Electrical Engineering, Technical University of Denmark
Simulation Results Storage usage and system performances
Storage Size Storage usage
(kWh)
Energy released (kWh)
Average hourly error
Max hourly error
Base case size 0% PV
0 0 17.3% >50.0%
5% 2.64 1.90 10.3% >50.0%
10% 4.45 3.54 7.2% 41.2%
15% 6.03 5.07 4.1% 19.8%
20% 7.43 6.46 1.5% 6.1%
25% 7.99 6.99 0.7% 2.6%
30% 8.10 7.10 0.6% 2.2%
50% 8.28 7.27 0.5% 1.4%
34 DTU Electrical Engineering, Technical University of Denmark
Conclusions and future developments
• The proposed work describes an energy management strategy for a 10 kW PV plant coupled with
a storage system.
• By knowing the meteorological forecast for the next 24h, the objective is to dispatch the PV
system and to be able to grant the scheduled hourly energy profile by a proper management of
the storage.
• The energy manager controls the storage in a predefined way in order to ensure that the hourly
energy production plan is respected, compensating the forecast errors and minimizing the storage
itself usage.
• The study is intended to provide also a methodology for the evaluation of the size of the storage
system in terms of power and energy.
• Further analysis will aim at extending the study along the whole year and at evaluating the
behaviour of the energy manager in case of different forecast horizons.
35 DTU Electrical Engineering, Technical University of Denmark
For further readings…
• M. Marinelli, F. Sossan, G. T. Costanzo, and H. W. Bindner, “Testing of a Predictive Control Strategy for Balancing
Renewable Sources in a Microgrid,” Sustainable Energy, IEEE Transactions on, vol.PP, no.99, pp.1-8, Jan. 2014.
• M. Marinelli, F. Sossan, F. Isleifsson, G.T. Costanzo, and H. W Bindner, “Day-Ahead Scheduling of a Photovoltaic
Plant by the Energy Management of a Storage System,” Universities Power Engineering Conference (UPEC), 2013
48th International, pp.1-6, Dublin, 2-5 Sep. 2013.
• F. Baccino, M. Marinelli, P. Nørgård, F. Silvestro, “Experimental testing procedures and dynamic model validation for
vanadium redox flow battery storage system,” Journal of Power Sources, Volume 254, 15 May 2014, Pages 277-286.
• F. Baccino, M. Marinelli, F. Silvestro, O. Camacho, F. Isleifsson, and P. Nørgård, “Experimental validation of control
strategies for a microgrid test facility including a storage system and renewable generation sets,” Integration of
Renewables into the Distribution Grid, CIRED 2012 Workshop, pp.1-4, 29-30 May 2012
• F. Baccino, S. Grillo, M. Marinelli, S. Massucco, and F. Silvestro, “Power and Energy Control Strategies for a
Vanadium Redox Flow Battery and Wind Farm Combined System,” Innovative Smart Grid Technologies (ISGT
Europe), 2011 2nd IEEE PES International Conference and Exhibition on, pp.1-8, 5-7 Dec. 2011.